Proposed issue: LocalAgentBridge 0.5.0.dev0 companion candidate — manual release testing#
Unpublished preparation draft. No tests were performed by generating these rows. Once posted, record results only in that issue; do not synchronize this draft as a second ledger.
- Release: 0.5.0.dev0 companion candidate, unreleased; LocalAgentBridge / OmniLAB workflow.
- Package:
localagentbridge-0.5.0.dev0-py3-none-any.whl. - Wheel SHA-256:
3fde801aa34b5769ed68185ed7dfd857825d819f3af807f49bcc798ce37bbb2c. - Sdist:
localagentbridge-0.5.0.dev0.tar.gz; SHA-2567afe70772012d09328d4cb38617a06a982fb10173cc2def9f2bab3efd7208f10. - Source base:
8b02eba3deb02cc24393421b317644b183114a09; qualified package source/index tree3a7d24cbf8c8d41c14d88fa6a6df326a6fa37ebb. - Pre-preparation staged tree:
43266f6c38b3cfa4edf99a7953d5d0b2c71db33e(qualification documentation added outside the frozen package export). HEAD alone does not identify the staged candidate. - Evidence origin: retained recovery/Windows qualification artifact manifests; artifacts were not rebuilt for this draft.
- Environment planned: Windows x64, Python 3.11+ with Tk, isolated kit/config/data, existing Ollama and exact installed GPT-OSS; exact Qwen tag retained. Actual Windows build, displays, devices and tester: not recorded.
- Checklist owner: nishdel (release owner; confirm assignment).
Test definitions: docs/MANUAL-RELEASE-TESTING.md on the development branch. Before opening the issue, replace /blob/companion-release-candidate/ in these links with the reviewed documentation commit and verify them. Product/package identity above stays distinct from the documentation commit. The integration commit is 91ee3a20bb869c36e23bb1c7813827de053637b3; the later documentation commit did not rebuild the retained packages. See source/package provenance. This candidate remains not release-qualified.
Summary — Passed: 0 / Failed: 0 / Blocked: 0 / Not tested: 12 / N/A: 0
Use this issue's table as the sole execution status authority. New rows are No / Not tested. Tested: No / Partially / Yes / N/A. Result: Not tested / Pass / Fail / Blocked / N/A. Issue registered: — / Not needed / Pending / linked issue number. Pass requires all expected results actually observed and Tested=Yes. Record the actual tester/date. Blocked/N/A need reasons; failed rows need Pending or a linked existing/new defect. A closed defect never passes its retest. N/A is excluded from passes.
On a candidate change, first preserve the previous table/evidence in an issue comment labelled with its candidate/hash. Record affected IDs and reset them to No / Not tested; retain unaffected results only with owner rationale and original candidate/tester/date. Use separate rows for separate environments. Keep the summary synchronized with current rows. Do not copy results into repository documents, infer human acceptance from automation, or open a new release-testing issue for each development task.
| ID | Test | Expected result | Tested | Result | Issue registered | Tester / date | Evidence / notes |
|---|---|---|---|---|---|---|---|
| LAB-M01 | Packaged launch | Dock opens from the identified package. | No | Not tested | — | — | — |
| LAB-M02 | Exact model selector | Exact tags appear; selected default persists. | No | Not tested | — | — | — |
| LAB-M03 | One local task | Saved local result awaits explicit review. | No | Not tested | — | — | — |
| LAB-M04 | Keyboard and focus | Controls work without refresh stealing focus. | No | Not tested | — | — | — |
| LAB-M05 | DPI and readability | Text, controls and evidence remain readable. | No | Not tested | — | — | — |
| LAB-M06 | Hide and reopen | Same run returns; dock exit leaves work independent. | No | Not tested | — | — | — |
| LAB-C01 | Conditional: missing Ollama | Unavailable inventory gives actionable guidance. | No | Not tested | — | — | — |
| LAB-C02 | Conditional: absent model | Exact missing tag is rejected without substitution. | No | Not tested | — | — | — |
| LAB-C03 | Conditional: workflow checkpoint | Saved handoff stops at an explicit decision. | No | Not tested | — | — | — |
| LAB-C04 | Conditional: live callback | One authorized review still needs a decision. | No | Not tested | — | — | Explicit callback approval required. |
| LAB-C05 | Conditional: clean Windows | Package starts on a genuinely clean host. | No | Not tested | — | — | Clean host must be supplied. |
| LAB-C06 | Conditional: perceived response | Bounded interaction meets predeclared criteria. | No | Not tested | — | — | Owner criteria required before scoring. |
Automated evidence (separate): Retained, not rerun here: baseline Windows suite 411 passed/1 privilege skip; installed recovery suite 50 passed; actual Python 3.11.9 suite 361 passed with symlink deselected and accepted recovery tests excluded. Native Tk and process-crash evidence do not pass human rows. See qualification and recovery.
Release-owner decisions (separate): Callback destination/payload/context, quota and retry limits require the existing private bounded approval request. No authenticated invocation is authorized by this issue. Fresh Windows, symlink privilege, reference hardware/workload and performance criteria remain separate owner decisions. Process-crash recovery is accepted within its documented limits; do not reopen that audit to fill manual gaps. No publication approval is implied. Preserve GPT-OSS preference and exact installed qwen3:14b-q4_K_M.
Candidate/retest history: initial documentation draft; no previous human results copied or overwritten. Preserve historical reports with their original scope. Future candidate changes require an explicit affected-ID list and the preservation procedure above.